Sharing Analytics Improves Outcomes, Revenue

This article appears in the September 2012 issue of HealthLeaders magazine.

In 1999, Microsoft chairman Bill Gates popularized the term "digital nervous system," which describes how technology could be used to let enterprises make better decisions faster, mimicking the autonomic nervous system of living organisms.

Today, leading healthcare organizations increasingly rely on their own digital nervous systems—improving the quality of care, sustaining operations, and ensuring profitability. With unprecedented transparency, the metrics of running a healthcare enterprise trigger rapid responses to all sorts of changing conditions.

Take, for instance, the Cleveland Clinic, a multispecialty academic medical center that handles 4.6 million patient visits a year. CFO Steven Glass knows that his organization requires constant awareness of key business indicators in order to stay on top of its game.

"I'm on the dashboards probably 320 days out of 365 days a year, and it's really a key way for me to pulse how the organization is actually performing," Glass says. "That includes volumes, occupancies, patient satisfaction, quality metrics, utilization, cost, all kinds of different classifications."

When Cleveland Clinic deployed its first dashboards eight years ago, they were "relatively basic, showing overall volume information in the organization," he says. "It's evolved drastically, to where now on our dashboards, we have information that's updated as frequently as every 30 minutes."

Why so often? "On a particular day we can be running at 95% occupancy, so we really need to understand that the house is full," Glass says. "It's critically important for executives across the organization. It is a way that, as the CFO, I know are we having a good month or a bad month.

"Long before you see revenue numbers and expense numbers, if you really know your business and you track it pretty closely, you can see where you've got trends in different parts of your organization," he says.

Glass says midway through the month, if he sees light volume in a unit, he will pick up the telephone to make sure expenses are being appropriately managed.

But the real key to today's business intelligence at an institution such as Cleveland Clinic is distributing the information gathered by the digital nervous system not just to top leadership, but to all those in the organization with a need to know.

"People know I'm going to call," Glass says. "They know executives are paying attention to this on a regular basis, so as a result, people aren't waiting for the phone call. It builds an expectation in front of people that this kind of information is being monitored, so people don't want to get that phone call and not have answers as to what's going on in their organization."


Single source of truth
Business intelligence dashboards don't just happen. Behind their creation is a vast amount of tech-powered preparation and careful design, as healthcare executives sift through the many sources of data being generated in their enterprises, looking for the holy grail: a commonly agreed–on set of key metrics aggregating data from widely different information systems throughout a modern hospital.

These rollups are commonly referred to by healthcare CIOs as the "single source of truth," which, when properly designed, also allow a wide variety of enabled professionals to drill down multiple levels from the top figures in dashboards all the way to performance of individual units and physicians, if needed, to understand trends.

Retail and financial enterprises learned early how to employ business intelligence to respond to customer demands quickly. Healthcare is a latecomer to this, driven by the more recent move away from pay-for-performance to value-based healthcare and accountable care.

In this new environment, "you're going to be judged on patient experience, quality outcomes, and how you are managing the cost for that value of service delivered," Glass says. "These dashboards are tools that are already in place that allow us to focus on that."

Analytics don't just let executives compare results over time, but they also let them measure performance compared to national metrics. "You want to be able to see how your docs are doing compared to other docs in the country, as well as your region; that includes length of stay, cost per case, utilization of pharmaceuticals, and mortality," says Rick Schooler, FACHE, FHIMSS, FCHIME, the CIO of Orlando Health, a 1,780-bed network that includes Orlando Regional Medical Center and five other hospitals.

About three years ago, Orlando Health realized it had to begin developing its own single source of truth, Schooler says. "We could not continue to have multiple metrics—often the same metrics coming from different resources and different sources—that reflect different answers for the same question," he says.

The goal was to populate an information repository to let service line workers go after whatever they need, Schooler says. "We've been heads down at this for maybe a year and three or four months, [and] we believe there is a three- to five-year initial deployment to get to where there is what we would consider a critical mass of information from across the organization."

Schooler selected an Oracle-based enterprise data warehouse hardware appliance from Columbus, Ohio–based Health Care Dataworks. "You basically then have to develop the feeds out of your systems to populate that data model and then you use different toolsets on the front end to give people access to the information and to slice and dice and to view and scorecard and dashboard yourself to your heart's content," he says.

Although many healthcare systems are new to data warehouses, Schooler isn't. As far back as 1988, he was implementing data warehouse technology in the telecommunications industry.

"When you think about the concepts of accountable care, and that many of us are going to end up managing a population of patients, multiple populations, and at some point we're all going to be assuming some level of risk to keep them well or risk to effectively manage the care we have to deliver, there's no way to survive without these kind of tools," Schooler says.

Uses of analytics
One of the key uses of analytics in healthcare is to improve the quality of clinical outcomes. At Allina Health, Chief Clinical Officer Penny Wheeler, MD, and her team used analytics to look at the timeliness of breast biopsies. "We could tell that some sites were doing same-day breast biopsies when a concern was raised with a worrisome breast lump, and some of our sites were doing it up to 10 business days later," Wheeler says. "That was able to guide an initiative and have us give oversight over timeliness of breast biopsies."

Analytics also revealed that 14% of the time, Allina Health clinicians were electively inducing labor before 39 weeks. It meant too many babies going to intensive care and too many cesarean sections, Wheeler says. In the past year, by putting analysis of this data in the hands of the clinicians who knew best, 235 women did not have to have C-sections, she adds.

As much preparation as these analytics systems require, they have become more automated than before. "Knowing that times are tough and knowing that more and more measures are being required of us, we couldn't laden the organization with hundreds of FTEs that were in the measurement and analysis area," Wheeler says. "Much of the business case for this was to actually rightsize the measurement and analysis team by creating automated platforms to get the information we needed."

With 11 hospitals and 82 clinics in Minnesota and western Wisconsin, Allina Health generates $3.5 billion in annual revenue. To build its current analytics system, Allina hired Healthcare Quality Catalyst in Salt Lake City, a consulting firm run by two former executives of Intermountain Healthcare.

Then Allina added another analytical layer using QlikView business intelligence software by Radnor, Pa.–based Qlik Technologies, Wheeler says. "It allows people to look at the data and interpret it without needing to wait four weeks for an analyst to do so," she says.

There have been measureable returns on Allina's investment, she says. For instance, Allina netted $1 million to $2 million in a pay-for-performance initiative from payers wishing to reduce unnecessary C-section costs.

"But remember, we've given up revenue to do that, when you're not doing as many cesarean sections," Wheeler says. "If you don't have this infrastructure and capabilities in place, then as we move to outcomes per dollar spent, you're kind of hosed." But with that infrastructure comes greater confidence of success because the leaders know how to focus their efforts.

Allina moved strongly into outcome-based payment this past January, when it kicked off its initial ACO efforts as one of 32 Pioneer ACOs in the United States, even working in conjunction with a rival provider. "We couldn't get to actionable data without the integrated data warehouse," Wheeler says.

Measuring key business indicators in a healthcare system often precedes clinical business intelligence, but not always. At Memorial Sloan-Kettering Cancer Center in New York City, use of a clinical data repository began 25 years ago, while the financial operational data warehouse is less than two years old.

A single hospital with 470 beds and a number of regional facilities, Memorial Sloan-Kettering Cancer Center evolved its clinical data warehouse six years ago to be Web-enabled and HIPAA-compliant, says Patricia Skarulis, vice president of information systems and CIO.

To help take care of patients closer to their homes, patients may have surgery at Memorial Sloan-Kettering but receive radiation at a regional facility in northern New Jersey, Long Island, or Westchester County in New York. "We have a database of over a million and a quarter patients, and we have everything about them for their inpatient and outpatient cancer care," Skarulis says.

Every event that occurs with those patients is time-stamped, allowing clinical analysis to begin quickly. "We can begin to simulate what a bottleneck might be," Skarulis says.

Memorial Sloan-Kettering's newly developed business intelligence warehouse, built on IBM's Cognos and SPSS software, can drive decisions from the detail-oriented work of matching invoices and purchase orders to more mundane analysis.

With budget planning for 2013 under way, unit managers will have the ability to spot changes such as how performance in a given month compares to the same month going back two years. Predictive analytics will still require expertise. "We have several analytics groups within the institution, including advanced analytics and predictive analysis to clinical work," Skarulis says. "They have been so used to doing it for all of their clinical studies that to be able to have this database that will match up square footage and buildings with people and salaries and expenses and all of that and revenues, to have a total comprehensive picture—it will be extraordinarily useful to some of these super-analysts to be able to pull together insights for our business operations."

Clinical implications
Jonathan Perlin, MD, is chief medical officer and president of the clinical and physician services group at HCA, a Nashville-based hospital company that includes about 163 hospitals and 109 freestanding surgery centers in 20 states and England. He notes that data can move the needle on best practices for healthcare.

"During the pandemic flu threat of a couple of years ago, we partnered with the CDC and the Department of HHS to share not just data about prescriptions for flu-related medications, but also to share something that we've been doing for a long time: anything that would indicate an increase in influenzalike illnesses, or laboratory data that would show increases in white blood cell count or other markers of infection," says Perlin, who is a former undersecretary of health for the Department of Veterans Affairs. "The ability to span multiple states, multiple markets, multiple environments, and detect trends and aggregate those data was immediately beneficial to the public health in terms of detection of potential trends in influenza."

More recently, HCA partnered with Richard Platt, MD, a hospital epidemiologist at Boston's Brigham and Women's Hospital, to help the Centers for Disease Control understand the best way to prevent MRSA infections in hospitals.

Pulling in data from 42 HCA hospitals, Platt and Perlin's team brought together information about infections, cultures and sensitivities, antibiotics, and changes in practice. The results, to be released in October, will suggest a definitive way to avoid MRSA infections in hospitals, especially for patients in intensive care, Perlin says.

"We conducted a study over 42 hospitals in three interventions to detect which was the best to reduce MRSA," Perlin said. "The benefit of large-scale information systems spread across a large number of hospitals is to study more than 70,000 patients in more than 240,000 patient days to identify definitively the best way to avoid MRSA in hospitals."

A third HCA study, still unreleased, analyzes different pain medications used in and around anesthesia. "We find that certain medications allow the patient to go home earlier than other medications," Perlin says. This study may yield not only clinician guidance for a better patient experience, but also allow HCA to deliver the best care as efficiently as possible, he adds.

"Good quality is good business," Perlin says. "We believe that good quality in healthcare has to be supported by robust information, supporting the day-to-day decisions with insight that derives from analytics."

The Joint Commission designated 76 of HCA's facilities as being among the 405 top performers in the country last year, Perlin says. "That's a disproportionately good representation in terms of the positive performance, and demonstration to the power to be able to use clinical analytics to drive focus on clinical performance and improve outcomes," he says.


Obstacles to analytics
The difficulties in implementing business intelligence mean that not every healthcare organization is able to jump right in. The scale of data being fed into today's healthcare data warehouses can be astronomical.  HCA attested to Stage 1 of meaningful use at virtually all of its 145 hospitals nationwide in 2011. HCA records more than one billion medication administrations annually, Perlin says.

"We want to put tools and information in the hands of clinicians and business leaders. We want to move as much as possible from a library metaphor to an Internet metaphor," Perlin says. By that, he means, "We want to move from something you have to go to a certain individual with a certain set of skills to answer every question to really provision users with the capacity to check out data and perform analytics, obviously in an appropriate, secure, and private fashion."

Complicating the move to these tools are migrations from point solutions that many healthcare systems have adopted under the influence of what Schooler calls "the tyranny of the urgent."

For years, technology providers have supplied analytics packages that explore one large but siloed system of information, such as revenue cycle, supply chain, or performance management of physicians.

"Ask the average organization to show you their enterprise data warehouse platform from which they do enterprise analytics on their data," Schooler says. "Ask someone to show you where that's all integrated in one spot, and you'll find few and they're far between.

"This is a discipline that our industry has to embrace and adopt and has to master going forward," says Schooler, who was named CIO of the Year for 2011 by the Healthcare Information and Management Systems Society and the College of Healthcare Information Management Executives.

Care management platforms, such as patient registries or disease registries, will also need to feed data into these data warehouses, Schooler says.

Another hurdle to implementing analytics is that all healthcare organizations also have to wrestle with both structured and unstructured data, says Jesse Spencer-Smith, director of clinical analytics with the clinical and physicians service group at HCA.

"Administrative data is coded and clear," Spencer-Smith says. "Data elements are highly structured. Some of our clinical data may be less structured and needs to be mapped. That's an ongoing project. Each of these presents its own challenges in terms of analyzing the data."

Schooler has some advice for hospitals trying to harness the power of analytics—particularly smaller community hospitals that lack funds to install expensive IT systems.

"Governance and prioritization are absolute keys," Schooler says. "As a standalone 250-bed community hospital, you may not need to invest in a separate platform. You may be running your entire organization off of one health information system."

Next, look for market partners, he says. For instance, two or three different vendors provide well-established physician performance technologies and can help providers fill gaps in their current data collection.

In a world where speed of insight is where healthcare providers keep their competitive edge, metrics that matter and the tools to create them will continue to provide that edge.

Reprint HLR0912-2

This article appears in the September 2012 issue of HealthLeaders magazine.